Deep reasoning, extended thinking for critical tasks

You send a simple prompt to Claude, it replies in 2 seconds. You send a complex case (math, code, planning), it replies just as fast, but the output is broken. Why? It did not have TIME to reason. Extended thinking is giving it that time explicitly, with a token budget dedicated to reflection before the answer. This course gives you the 4 foundations: (1) when thinking really pays off vs when it wastes, (2) how to parameterize a thinking call (budget, output thinking block + text block), (3) the thinking + tool use combo that changes agent quality, (4) the cost/quality arbitrage to enable thinking at the right moment in prod. By the end, you know how to arbitrate between speed and depth, and you build critical features (debug, planning, decision trees) you did not dare build before.

Level : intermediate

Modules

  1. 1. Why thinking, when to ask Claude to reason
  2. 2. Anatomy of a thinking call, parameters, budget, output
  3. 3. Thinking + tool use, the combo that changes everything
  4. 4. Cost/quality arbitrage, when to enable in prod

View on Coeurdar